Detection Is Not Early Warning: Evaluating Token-Level Alarms on Language Model Failures
cs.LG, cs.CL
Submitted: 2026-11-12
Updated: 2026-11-13
Code: https://github.com/swiss-ai/degeneration-probe
Terminology
Sources
- Language Models Can Predict Their Own Behavior
- When Evidence is Sparse: Weakly Supervised Early Failure Alerting in Dialogs and LLM-Agent Trajectories
- Obfuscated Activations Bypass LLM Latent-Space Defenses
- Llama-Nemotron: Efficient Reasoning Models
- HuatuoGPT-o1, Towards Medical Complex Reasoning with LLMs
- Circular Reasoning: Understanding Self-Reinforcing Loops in Large Reasoning Models
- Real-Time Detection and Repair of LLM Agent Failures
- From Probe Scores to Alarm Policies: Operational Validity of Activation Monitors for Language-Model Agents
- Detecting Strategic Deception Using Linear Probes
- The Llama 3 Herd of Models
- DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoning
- To Steer or Not to Steer? Mechanistic Error Reduction with Abstention for Language Models
- PrefixGuard: Online Failure Warning and Trace-Grounded Diagnosis for LLM Agents
- Quickest Detection of Hallucination Onset: Delay Bounds and Learned CUSUM Statistics
- Mistral 7B
- Predict, Don't React: Value-Based Safety Forecasting for LLM Streaming
- Programming Refusal with Conditional Activation Steering
- From Judgment to Interference: Early Stopping LLM Harmful Outputs via Streaming Content Monitoring
- Detecting High-Stakes Interactions with Activation Probes
- Real-Time Detection of Hallucinated Entities in Long-Form Generation
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis